There are two types of AI used in marketing and they are constantly confused. The confusion is not semantic: it leads to incorrect investment decisions. This distinction matters.
Generative AI creates new content: text, images, video, audio, code. Language models (Claude, GPT-4,LLaMA) are generative AI. Image generators (FLUX, Midjourney) are generative AI. When you use AI to write an email, generate ad copy or create a product image, you are using generative AI.
Predictive AI analyses historical data to predict future behaviour. Lead scoring models, churn prediction systems, predictive email that calculates the optimal send time, and the recommendation system of an e-commerce site: all of that is predictive AI. It does not create anything new: it analyses patterns in data to anticipate what is going to happen.
When to use generative AI in marketing
Generative AI has the most impact where the main bottleneck is production: generating enough quality content to feed channels, producing enough creative variants for tests, personalising messages at scale.
The clearest use cases:
- SEO content production: articles, product descriptions, metadata, FAQ pages
- Campaign copy: ad variants, nurturing emails, video scripts
- Message personalisation: emails generated in real time based on the recipient’s context
- Visual assets: product images, social media banners, ad creatives
What generative AI cannot do: make decisions based on historical data, predict behaviours, or automatically optimise the budget. For that, you need predictive AI.
When to use predictive AI in marketing
Predictive AI has the greatest impact where the main challenge is making better decisions with the data available: who we send which email to, who we show which advert to, which leads the sales team prioritises, when the right moment is to make a renewal offer.
The clearest use cases:
- Lead scoring: who is most likely to buy in the next 30 days?
- Churn prediction: which customers are most likely not to renew?
- Predictive email: when is each user most likely to open an email?
- Predictive audiences in ads: which users in the inventory are most similar to my best customers?
- Attribution: which channels genuinely contribute to conversions?
What predictive AI cannot do: generate email content, create ad creative, write the blog article.
The full stack: generative + predictive working together
The greatest impact comes when the two work together. The predictive model identifies who to send the message to (lead scoring, predictive segmentation). The generative model creates the personalised message for that profile (real-time text generation). The predictive model determines the optimal moment to send it (open window). The predictive model measures the impact and feeds the system back for the next iteration.
That is what a well-designed AI marketing system does: it is not just generative AI to produce more content, nor just predictive AI to optimise distribution. It is both working together.
In our article on marketing with AI: complete guide we covered how to integrate both types into a coherent strategy.